Genetic algorithm for humanoid walking online on Intellect

Used 520 times

Genetic algorithm for humanoid walking training

The simulator trains a population of simple humanoids to walk forward. After each round, the best genomes are copied, crossed over and mutated, so over generations the gait usually becomes more stable and effective.

Generation: 1 Best score: 0.00 Current leader: - Tick: 0

Controls

Gene mutation probability Gene mutation amount Champions to copy Round length Animation quality Simulation speed

Current generation

Name	Reached	Stability

Best results

Gene	Name	Reached

Description

What is it?

This is an educational genetic algorithm simulation. Movement is built physically: the body consists of connected masses, while muscles change joints and interact with gravity, friction and the ground. The genome defines gait parameters: step amplitude, foot lift, knee bend, torso lean and movement rhythm. The model does not use external libraries, but keeps the main mechanics of the original: the population competes, the best individuals produce offspring, and mutation explores new variants.

How is the score calculated?

The score increases for moving forward, keeping head height even, alternating steps and maintaining a stable body position. Falling, a head that is too low and chaotic walking reduce the final result.

Was this answer useful?

Choose a quick rating so we can improve the next answer for you.

How satisfied are you?

Your answer option for this service or noticed an error:

This interactive model shows how a genetic algorithm can gradually tune a humanoid walking pattern. Motion is calculated by an embedded physics model: point masses, gravity, friction, foot-ground contact, bone-length constraints and motorized joints. Each population member has genes for stride length, foot lift, body lean, knee bend, movement period and phase shift. After every round the best humanoids have a higher chance to pass their parameters to the next generation, while mutation adds random changes.

Algorithms Programming